OpenAI's $5M Astra Defense Beaten by 3 Guys with $5K of Opus, Argues Viral Thread
harris_edouard · x · 2026-09-19
- a16z quoted Greg Brockman saying OpenAI pulled 25% of its production engineers to red-team its own systems with Astra, claiming all P0 vulnerabilities Astra could find were fixed, with new rounds planned per model.
- @Laneless countered that this defense stack — Astra plus a quarter of OpenAI's production engineers plus a conservative $5M in inference — was defeated by roughly three people with $5K of Claude Opus inference on offense, concluding everything might be cooked.
- The debate highlights the stark offense-defense cost asymmetry in AI-era cybersecurity.
Related event: OpenAI Pulled 25% of Engineers to Hunt Bugs with Its Astra Model(4 posts)→
More from Safety
- Binance breach criticized: DPRK fake-recruiter malware tactics were publicly documented for years — AccBalanced · 2026-11-04
- OffsecCyber dataset combines HackerOne reports and CVEs to train pentesting LLMs — AccBalanced · 2026-11-04
- Abliterated GLM-5.3 pulls off red-team exploit in 3:41 for $0.02, stock model refuses — bratton · 2026-11-04
- AI industry bet on both parties: $55M+ spent this election cycle via dual PACs — ChrisGPT · 2026-11-04
- Valid tokens, allowed actions, zero intent checks: the agent security gap nobody solves — Cytersmyz · 2026-11-04
- Princeton paper: harsher penalties can make AI auditors inspect less — Few-Decision6287 · 2026-11-04